Subproject 2: Numerical methods for optimization-based control of cyclic processes
The working group for systems theory, control and optimization at the University of Freiburg has many years of experience in the development of efficient numerical methods and software for nonlinear model predictive control and their application to systems with time scales in the milli- and microsecond range. In particular, the open-source software packages BLASFEO and acados are developed within the group.
BLASFEO provides high-performance linear algebra operations customized for embedded optimization, which are successfully used in the software package acados, among others. The software package acados provides efficient algorithms for solving optimal control problems on embedded hardware and has already been successfully applied within the research project.
Goals in the second funding period
The goal of this subproject is to enable real-time control of fast cyclic processes using innovative algorithmic ideas. The challenges associated with real-world operation, in particular of the PCCI and GCAI engines in Aachen and Zurich, call for new algorithmic developments and numerically favorable problem formulations.
In the second funding period, the following projects in particularly are to be worked on:
- Parallel online optimization based on a multi-level architecture
- Interleaving of optimization problem formulations and numeric
- Approximate methods for robust control of fast cyclic processes
Highlights of the first funding period
- Discrete model formulation for real-time control of the GCAI process
- Zero-order optimization approach
- Inexact Newton with Iterated Sensitivities (INIS)
- Development of the open-source software acados
Univ.-Prof. Dr. rer. nat. Moritz Diehl